Artificial intelligence in Healthcare
summary
A production system called mosaic PV ingests diverse sources (PDFs, transcripts, calls), runs parallel NER engines plus rule‑based, statistical and deep learning models, auto‑codes to MedDRA, and surfaces recommendations with a human-in-loop audit trail. It processes ~250,000 adverse events a year (growth ~10%/yr), flags the 10% novel cases, and reduces end‑to‑end case handling from hours to seconds via an AWS serverless stack (Lambda, Step Functions, Transcribe Medical, SageMaker) with model validation and regulatory documentation.
AI-written summary · about
chapters (18)
- 0:00Introduction
- 1:58PwCs AI approach
- 10:38Roadmap
- 13:14Conceptual Architecture
- 17:56Bias
- 19:32Handwritten notes
- 21:13Key challenges
- 23:26AI vs PA
- 26:43Validation
- 27:50Manual intervention guidance
- 28:53Handwriting
- 29:28Numbers
- 31:31Lessons Learned
- 34:08Preconceived notions
- 35:43AI Machine Learning
- 44:59Show of Hands
- 48:10Deep Dive
- 51:50Architecture
related talks
09mHow AI is changing the healthcare system
05mHow generative AI can support medical staff
47mThe Future of Healthcare and AI
31mAI For Good: Fighting Health Insurance with AI
21mTransforming Healthcare Through Data-Driven Intelligence
30mRevolutionising epilepsy care with AI/ML
16mGenAI in Healthcare Claims: Processing and Detecting Fraud
45mGenerative AI in medicine: Improving patient & provider experiences